Posts for ntclark

Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
I'd definitely be more interested in seeing all the missions run as fast as possible rather than skipping some.
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
Great to hear! Nice WIP, and I really hope the motivation stays with you!
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
Having played through the game recently, I enjoyed it a lot. Agree with Sir VG though that the character was ridiculously overleveled. The MP limitation of a lower level would break up the constant warping and be much more impressive from a resource management stand point.
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
FWIW, I'm really enjoying reading about the planning and watching the WIPs for this run. Can't wait for the final product!
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
God I f'ing hate nicovideo...
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
Seriously, Kriole, your WIPs are GREAT. Keep it up. I'm sure there are many of us lurkers following the thread enjoying your progress.
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
Pretty much every recent 32-bit x86 chip can have 64 GB of usable memory even though each 32-bit application can only address 4 GB. Source: http://en.wikipedia.org/wiki/Physical_Address_Extension
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
If you're asking what do you consider 100%, I would say no on both. The 100% category in RPGs usually doesn't include killing all unique monsters or leveling up to the maximum extent possible. If you're asking from an entertainment perspective, I'd enjoy seeing the bosses killed, but don't particularly care what level you reach.
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
Having struggled to beat this game, I really enjoyed the movie. Although I could see how someone who hadn't played the game might be bored, I still think the quality is there, thus a 'yes' from me. Nice job!
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
Are you sure that the optimal solution for an euclidean TSP can be found in polynomial time? Or simply a solution which is "good enough"? Remember that we are talking about absolute optimal solutions here.
No, I'm not sure about euclidian TSP, and I probably shouldn't have used that example. Here's an example I am sure about: many common instances of integer linear programming (an NP-Hard problem) reduce to the max network flow problem, which can be solved optimally in polynomial time. Thus good ILP solvers will check for this special case and solve it optimally very quickly. I assume there are similar TSP formulations which can also be solved optimally without having to check an exponential number of solutions.
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
The TSP itself is already in the complexity class NP-hard, which means that there's no algorithm for solving it in a reasonable amount of time for all but the smallest values.
Not exactly correct. NP-hard means there's no known algorithm for solving all possible inputs in a reasonable amount of time. There are always special cases (e.g., euclidian tsp) that can be solved efficiently on large inputs in a reasonable amount of time, even on conventional computing devices.
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
actually TASing flash games is likely considerably easier to implement than general windows games, and you wouldn't have to turn them into .exe files. That's because there is already an open source virtual machine for executing flash applications: http://en.wikipedia.org/wiki/Tamarin_%28JIT%29 Someone would simply have to add the appropriate hooks for TASing (e.g., frame advance, etc.).
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
Looking good!
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
Flygon, that runs really slow (~5fps) for me. Same as on the viddler site (i.e., non-embedded).
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
Here's a brute force bot I made for SNES Super Punchout. http://www.cc.gatech.edu/~ntclark/software/punchout_brute_force.lua It uses a branch-and-bound algorithm, with very game-specific bounding. I suspect any brute force technique will have to be very game specific. I know the code is ugly, but I was just playing around. The current implementation uses a save-state before every new input, and then rolls back to the save states to enter new input combinations. This was stupid; saving the state takes way too long (have to write to disk), and if I had to redo it I would create one state at the beginning and just apply all the different move orders to that initial state.
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
Thank you for making such a great run!
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
cool. thanks for the help!
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
i got a desync during the first mission. anything special i need to know about your settings?
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
Mweah, with snes9x.speedmode('maximum') my 2 year old notebook can do around 1200 frames/sec. When you've got something of say 20 frames, and you want to use 4 buttons (say right/up/jump/speed), you got . So Saturn is right, that was probably the most useful bruteforcing we will ever have for this game
Be careful with how quickly you write off brute forcing. There are many important problems where the best known solutions are essentially brute force, and many, many really smart people have developed ways to effectively solve them in many cases. Techniques like branch and bound and dynamic programming. The simple calculation 4^20 ignores these techniques and any domain specific knowledge which can be used to quickly prune unpromising search paths. For example, if you're looking for the optimal path through a room in super metroid you can stop searching any paths where your HP drops if you know there's no useful damage boost in that particular room. Or if you're trying to brute force the kraid fight, there's no need to search paths which fire shots when kraid is still underground. I'm not saying the whole game will ever be brute forced, but certainly it can do more than 40 frame segments.
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
Ah, sorry about that. I was confused about what you were asking. GG -> PAR conversion: http://www.videogamesource.com/genie/gg/gghandbk/p12.html
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
The PAR codes are just an address followed by a value, so it's pretty easy to turn the addresses in the script into PAR codes (and vice-versa). For example, if you wanted the opponent to always be hittable in the top-right (address 7e096b) use the PAR code 7e096ba0, where the the last two digits, a0, are 160 (unblocked) in hex.
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
Hi all, I was bored last week and started messing around with Lua scripting, trying to write a tool to brute force super punch out fights. It isn't quite fast enough at the moment, but I'm bored working on it, and thought someone else might find want to play with it or improve it (yes, I know the code is atrocious. I just wanted to get something working fast): http://www.cc.gatech.edu/~ntclark/software/punchout_brute_force.lua If you've got a fast CPU and enough time on your hands, you could probably prove some of the shorter fights (e.g., piston hurricane) are optimal in Phil's movie. But even if you couldn't care less about lua, there are quite a few useful memory addresses documented in the script, which I haven't seen collected in any other location.
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
Minor stylistic suggestion: perhaps shooting in some interesting pattern (e.g., with the music) would make the long walks a bit more entertaining.
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
Raiscan wrote:
Too bad that whomever added that revision forgot to add it was tool-assisted, and not just done by someone on an emulator..
Being that this is Wikipedia that should be pretty easy to change, no? I mean, hell, we could change the article to say that it was played on an emulator by a legion of trained ninja-monkeys, or something even less believable.
Experienced Forum User
Joined: 5/9/2008
Posts: 35
Location: USA
The sodomy after the second Aganhim fight made me laugh out loud : )